On Tuesday, 5 June 2012 at 13:27:12 UTC, Roman D. Boiko wrote:
maybe it is better to avoid immutability... or do bulk ins /del
before copy.
Would it be difficult to introduce two optimizations:
* have strings of all lengths above configurable threshold go to
the same bucket (is it reasonable?)
* have ability to store a checkpoint, then do a series of
modifications, and store another checkpoint but reuse data which
have not been affected.
For the second optimization a possible design would be to store
internal state as snapshot + diffs, and apply diffs when creating
another snapshot. Diff format should allow efficiently performing
trie interface operations.
